改进的B-P神经网络系统在储层敏感性伤害预测中的应用

从储层保护的角度出发,对长庆油田的储层伤害进行了综合分析研究,确定了对储层造成伤害的几种因素:水敏、盐敏、速敏、酸敏、碱敏,并采用Kohonen自组织网络和改进的B-P网络相结合的组合神经网络技术建立了储层敏感性伤害的预测模型.该模型改进了以往神经网络模型在数据处理方面的缺点,缩短了网络学习训练的时间.运用该模型对长庆油田储层伤害进行了预测,预测结果与实测结果较一致性,说明改进后的神经网络模型在储层敏感性伤害预测中能够满足工程预测的需要,从而为油气层保护技术措施提供可靠的依据.

Application of modified B - P neural network system in reservoir sensitivity damage prediction

From the perspective of reservoir protection,an all-around analysis has been conducted for reservoir damage in Changqing Oilfield,and several factors which result in reservoir damage are confirmed: water sensitivity,salt sensitivity,rate sensitivity,acid sensitivity and alkali sensitivity.With the help of Kohonen self-organized network and modified B-P network techniques,a prediction model of reservoir sensitivity damage is established.This model overcomes the shortcoming of former neural network models in terms of data processing,and also lessens the time of network learning and training.The model is used to predict reservoir damage in Changqing Oilfield,the prediction result is quite consistent with the measured result,which verifies that the improved neural network model is reliable in prediction of reservoir sensitivity damage,and the model can meet the requirement of engineering prediction,thus it can provide reliable basis to reservoir protection.